Action Spectra in Semiconductor Photocatalysis

Lee, S-K., Mills, A., & O'Rourke, C. (2017). Action Spectra in Semiconductor Photocatalysis. Chemical SocietyReviews, 46, 4877-4984. https://doi.org/10.1039/c7cs00136c

Published in:Chemical Society Reviews

Document Version:Peer reviewed version

Queen's University Belfast - Research Portal:Link to publication record in Queen's University Belfast Research Portal

Publisher rights© The Royal Society of Chemistry 2017This work is made available online in accordance with the publisher’s policies. Please refer to any applicable terms of use of the publisher.

General rightsCopyright for the publications made accessible via the Queen's University Belfast Research Portal is retained by the author(s) and / or othercopyright owners and it is a condition of accessing these publications that users recognise and abide by the legal requirements associatedwith these rights.

Take down policyThe Research Portal is Queen's institutional repository that provides access to Queen's research output. Every effort has been made toensure that content in the Research Portal does not infringe any person's rights, or applicable UK laws. If you discover content in theResearch Portal that you believe breaches copyright or violates any law, please contact openaccess@qub.ac.uk.

Download date:15. Aug. 2020

1

Action Spectra in Semiconductor Photocatalysis

Soo-Keun Leea, Andrew Mills*b and Christopher O'Rourkeb

a: DGIST, 333, Techno Jungang Daero, Hyeonpung-Myeon, Dalseong-Gun, Daegu, 42988,

Korea; b: School of Chemistry & Chemical Engineering, Queen´s University Belfast,

Stranmillis Road BT9 5AG, Belfast, United Kingdom.

E-mail addresses: andrew.mills@qub.ac.uk

Abstract

Action spectra are an increasingly important part of semiconductor photocatalyst research,

and comprise a plot of photonic efficiency, η, versus excitation wavelength, λ. The features

and theory behind an ideal photocatalytic system are discussed, and used to identify: (i) the

key aspect of an ideal action spectrum, namely: it is a plot of η vs λ which has the same

shape as that of the fraction of radiation absorbed by the semiconductor photocatalyst, f,

versus λ and (ii) the key requirement when running an action spectrum, namely, that the

initial rate of the photocatalytic process is directly proportional to incident photo flux, ρ, at

wavelengths where η > 0. The Pt/TiO2/MeOH system is highlighted as an example of a

photosystem that yields an ideal action spectrum. Most photocatalytic systems exhibit non-

ideal action spectra, mostly due to one or more of the following: light intensity effects,

crystal phase effects, dye-sensitisation, dye photolysis, charge transfer complex, CTC,

formation and localized surface plasmon radiation, LSPR, absorption by a deposited noble

metal catalyst. Each of these effects is illustrated using examples taken from the literatures

and discussed. A suggested typical protocol for recording the action spectrum and

absorption/diffuse reflectance spectrum of a photocatalytic system is described. The

dangers of using a dye to probe the activity of a photocatalysts are also discussed, and a

possible way to avoid this, via reductive photocatalysis, is suggested.

Key words: photocatalysis; action spectrum, titanium dioxide; photosensitisation;

photolysis; spectral sensitisation

2

1. Introduction

Semiconductor photocatalysis continues to be an area of significant growth both in terms of

research and commercial products1. The latter are largely dominated by architectural

materials, usually accompanied by the sobriquet 'self-cleaning', such as self-cleaning glass,

tiles, paint and concrete1. In their 'self-cleaning' role, most commercial photocatalytic

materials use, as the semiconductor photocatalyst, TiO2, usually in anatase form, to mediate

the oxidation of pollutants, P, usually organic in nature, by ambient oxygen, i.e.

TiO2 P + O2 → oxidation products (1) UV

Note: the absorption of UV radiation is necessary to activate the TiO2, since it has a bandgap

of 3.0 eV.

Most commercial photocatalytic materials are slow to act, not least because only about 4%

of solar light is UV radiation, and most internal light sources have only a small UV

component, if at all. It is no surprise, therefore, to note that many research groups are

focused on developing a robust, visible light-absorbing semiconductor photocatalyst.

Indeed, a brief inspection of the literature1 reveals such a plethora of visible-light absorbing

photocatalysts, not least those based on anion-doping (N, S and P) of TiO2, that it is

surprising that, to date, that most major commercial photocatalytic products, such as self-

cleaning glass or tiles, remain TiO2-based and so only UV absorbing.

There are many possible reasons why current visible-light promoted photocatalysts

reported in the research literature have not made a significant impact on the major

commercial, self-cleaning product scene, including: the cost of scaled-up production, low

photostability and low activity, both in the visible and UV. However, another possible

reason is that the original claim of visible light photocatalysis has been made in error,

particularly if it was based on the photobleaching of a dye, as the test pollutant, so that

some or all of the observed photobleaching may have been due to dye photosensitisation

and/or dye photolysis, rather than photocatalysis2,3. Alternatively, if the pollutant forms a

visible-light absorbing ligand-to-metal charge transfer complex, i.e. LMCTC, or CTC for short,

with the semiconductor, then the visible light induced disappearance of the organic

pollutant may be due to a dye-photosensitisation mechanism, involving the electronically

3

excited state of the CTC, and, once again, not photocatalysis4. Given this range of possible

routes to the destruction of 'P' in reaction (1), it is obviously, essential to establish the

identity of the light-absorbing species (e.g. semiconductor, dye or CTC?) before any claim of

visible light photocatalysis can be made with confidence; and, as we shall see, this is usually

made possible by recording the action spectrum of the system.

2. Dye-sensitisation and photolysis

Dye (photo)sensitisation usually involves the initial electronic excitation of the dye, D, to D*,

by light of sufficient energy, hν', so that D* is able to then inject an electron into the

conduction band of the semiconductor, SC, which in turn then reacts with ambient O2

present in the system4,5. The process can be summarized as follows:

hν' SC O2

D → D* → D+• + SC(e-) → D+• + O2- (2)

where hν' is the energy of the photons absorbed by the dye, SC is the semiconductor under

test (usually TiO2), and D+• is an oxidized radical of the dye that is unstable and able to react

further to produce colourless degradation products. An example of a deliberate study of

dye sensitization is provided the work of Wu et al.5, in their work on the photosensitized

bleaching of Rhodamine B under visible light in an aqueous dispersion of TiO2, and

numerous reviews2,3 provide many more such examples. As noted above, a slightly

modified version of the dye-sensitisation process can also operate if the non-visible light

absorbing pollutant, P, forms a visible-light absorbing CTC between the SC and the substrate6.

Dye photolysis involves the electronically excited state of the dye which is either simply

unstable and/or quenched by ambient O2 to produce singlet oxygen – which then oxidises

the dye, i.e.

D + hν' → D* (+ O2). → bleached products (3)

Whatever the underlying mechanism, in dye photolysis the dye is bleached under

irradiation, often at a rate that is largely independent of the presence of the semiconductor.

In practice, in most photocatalytic studies the process of dye photolysis is usually quickly

discounted by observing no dye photobleaching in the absence of the semiconductor

4

photocatalyst; although, as rightly noted by others, this latter 'blank' experiment doesn't

eliminate the possibility of dye-sensitisation3.

A typical example of dye photolysis is the reported irreversible photobleaching of resazurin,

Rz, in an anaerobic aqueous solution containing P25 TiO2 and glycerol6. Rz absorb little in

the UVA region (320-400 nm) and so UVA irradiation of the Rz/TiO2/glycerol system results

in the rapid (half-life, t1/2, = ca. 1 min; 5 mW cm-2) photocatalysed reduction of Rz (blue

colour) to resorufin (pink), and concomitant oxidation of glycerol. In contrast, when

irradiated with 617 nm light (the λ(max) of Rz = 658 nm), where Rz does absorb, no Rf is

produced, but the blue coloured dye slowly bleaches (t1/2, = ca. 30 min; 10 mW cm-2) due to

dye photolysis; as might be expected, the rate of the latter process is independent of the

presence of glycerol and TiO2.

3. Action spectrum and photonic efficiency

In semiconductor photocatalysis an action spectrum is a plot of photonic efficiency, not

quantum yield7, versus excitation wavelength, λ, where photonic efficiency, η, is defined as

the ratio between the number of molecules formed or degraded in a photocatalytic system

per unit time, and the number of photons incident on the system per unit time, at a given

wavelength, λ8. In most cases, initial rate, ri (units: moles s-1) is taken as the numerator of

this ratio and the incident photon flux, ρ (units: moles of photons of wavelength, λ, or

Einsteins s-1) as the denominator, i.e.

η = ri/ρ (4)

In most photocatalytic studies the irradiance, E (units of mW cm-2), rather that ρ is reported,

where, for example, 1 mW cm-2 of 365 nm UVA light is equivalent to 1.84x1015 photons cm-

2s-1, or 3.05x10-9 Einsteins cm-2 s-1. If polychromatic radiation is used, then the ratio of initial

rate to incident photon flux has been defined8,9 as the formal (or apparent) quantum

efficiency, FQE, although many use the two terms, η and FQE, interchangeably9. In contrast

to photonic efficiency, quantum yield, QY, measurements are notoriously difficult to make10,

as they require the accurate measurement of the absorbed, rather than incident, light flux.

Thus, the measurement of QY, and the much less common reports of QY vs λ are outside

the scope of this tutorial review, except where such plots appear wrongly named, and are

really η vs λ plots, vide infra.

5

It is worth deviating slightly at this point to ask what – in the context of eqn (4) – is

monochromatic (rather than polychromatic) radiation. A brief perusal of the literature

associated with action spectra, and the measurement of η, suggests that it is not unusual

for an excitation emission band with a full-width at half-maximum intensity (FWHM) of ca.

20 nm, to be considered sufficiently monochromatic to allow the calculation of η; although

FWHM values nearer 10 nm are more common, vide infra. In order to achieve this level of

monochromaticity when recording an action spectrum, i.e. a plot of η vs λ, usually a

polychromatic light source, such as a Xe lamp, is coupled to a monochromator (MC).

Alternatively, interference-type optical filters maybe used, instead of a monochromator, to

provide different wavelengths of 'monochromatic' radiation. However, optical cut-off

filters, instead of interference filters, are not so useful, since they provide an integrated

version of the action spectrum, i.e. a 'pseudo-action spectrum'3, and although it is possible

to simply differentiate a pseudo action spectrum, to obtain the real action spectrum, it is

not recommended, since the latter may be significantly distorted due to a variation in the

sharpness of the filters and light intensity effects (vide infra).

Interestingly, the standard Black Light Blue lamp, BLB, with a phosphor which emits at 368

nm, and which is in general use in semiconductor photocatalyst research, has, typically, a

FWHM of ca. 18 nm. In contrast, the other main type of BLB, with a phosphor that emits at

352 nm, has, a FWHM of ca. 40 nm. The narrow band BLB is of little use with regard to

recording an action spectrum, but can provide a useful guide to the intensity dependence of

the photocatalytic system at 365 nm, and so be used to test the validity of the often made

assumption that the rate of the photocatalytic reaction is proportional to ρ (vide infra). For

reasons that will become clear later, ideally, whatever the irradiation system, the irradiance

should, ideally, be adjusted, using neutral density filters or by altering the distance, so that

the number of photons incident to the reaction cell is the same at all wavelengths. In

practice, this is a tedious process and so rarely carried out.

Obviously, if either dye photolysis or dye-sensitisation occurs, then a photocatalyst can

appear to exhibit visible light photocatalytic activity, whereas in fact the observed

photobleaching of the dye is NOT associated with semiconductor photocatalysis, i.e. NOT

due to reaction (1), but instead is due to a dye-sensitisation and/or dye photolysis

processes, i.e. reactions (2) or (3). As a consequence, in order to support any claim of

6

photocatalysis, especially visible light photocatalysis, there is a real need to identify the

absorbing species and this can be achieved by recording the action spectrum of the system,

which is usually a plot of photonic efficiency (or apparent quantum efficiency), η, of the system

versus wavelength of the incident irradiation3,11. If the action spectrum has a similar spectral

profile to the photoabsorption spectrum of the semiconductor alone, then this can be taken

as strong evidence that the reaction is indeed photocatalytic. If, on the other hand, there

are striking discrepancies between the two then this suggests a non-photocatalytic

mechanism is in operation.

4. Examples of action spectra

A list of some of the action spectra reported to date for both UV and visible light absorbing

photocatalysts is given in table 18,12-39, a brief examination of which reveals that most

involve the semiconductor photocatalyst TiO2 for UV-absorbing photocatalysts. Action

spectra for visible light absorbing photocatalysts include: CdS, WO3 and non-metal-doped

TiO2. Most action spectra can be identified as being a member of one of the following two

groups: (i) ideal (rare) or (ii) distorted (common) action spectra. An ideal action spectrum is

one in which the initial rate of the reaction is proportional to ρ at all λ, and the

semiconductor photocatalyst is the only absorbing species, so that the resulting action

spectrum has the same spectral profile as the plot of the fraction of light absorbed by the

semiconductor photocatalyst, f, versus λ. All photocatalyst action spectra that do not

satisfy the definition of the ideal action spectrum, fall under the heading of distorted action

spectra. Further discussion of ideal action spectra and the different types of distorted

action spectra is provided in the following sections.

7

Table 1 Examples of reported photocatalyst action spectra

Semiconductor Reaction Comments (FWHM/nm)* Ref

UV photocatalysis

TiO2 Stearic acid Similar action spectrum to that reported for acetic acid (7.5)

12

P25 Phenol, 4CP Xe/Hg lamp (10) 8

TiO2 H2 evol. and Ag dep.

5 different types of TiO2 usede for the 3 different photocatalytic reactions (23)

13

P25 Phenol, 4CP Intermediate analysis, Xe/Hg lamp (10) 14

P25 4CP Xe lamp 15

10% NaOH coated Rh-TiO2

H2O splitting Relative quantum efficiency gradually increases with decreasing wavelength

16

Pt-TiO2 H2O splitting Platinised anatase TiO2(BDH) used for the reduction of H2O. rate vs wavelength

17

TiO2 sol Salicylate Quantum yield vs excess photon energy 18

Visible photocatalysis

P25, Pt-WO3 Acetic acid Action spectrum similar to abs. spectrum 19

S-TiO2 Acetic acid P25<S-TiO2<1wt% Fe2O3 S-TiO2 20

N-TiO2 Phenol, 4CP Xe/Hg lamp (10) 21

Au-TiO2 2-propanol Action spectrum similar to DRS (15) 22

C3N4 H2O splitting band path filter (± 20 nm) 23

TiO2, S-TiO2 MB, acetic acid

Dye sensitised reaction was found with MB degradation (17)

24

CdS MB N-demethylation of MB accelerated only by the excitation of CdS in aqueous solution.(10)

25

Pt-CdS H2O splitting Only light of λ< 560 nm produces H2 and yields increased sharply close to the semiconductor band edge (2.4 eV, 510 nm)

26

*unless stated otherwise, a Xe-lamp with monochromator was used as the excitation source.

8

5. Ideal action spectra

In an 'ideal' heterogeneous semiconductor photocatalytic system, it is assumed that the

initial rate of a photoreaction is proportional to the rate of absorption of the exciting light,

so that eqn (4) transforms into:

η = (kfρ)/ρ = kf (5)

Where k is a proportionality constant, which, amongst other things, depends upon how

efficient the photocatalyst is at promoting the photocatalytic reaction. The term, f,

represents the fraction of exciting radiation of wavelength, λ, absorbed by the

semiconductor photocatalyst. The conventional band theory view used to rationalize

semiconductor photocatalysis suggests that the excess energy of any photons with energy >

the bandgap of the semiconductor is thermalized and, as a result, k should be largely

independent of λ8. If k is independent of λ, then a plot of η vs λ, i.e. an action spectrum of

the photocatalytic system, should simply reflect the extent to which the semiconductor

absorbs the light and so should have the same spectral profile as f vs λ. A rough

approximation of the shape of the latter profile is provided by the absorption spectrum, or

diffuse reflectance spectrum, i.e. DRS, of the photocatalyst3,9.

Diffuse Reflectance Spectroscopy, DRS, is used routinely to identify wavelength the region

where a solid material absorbs by measuring the reflectance, R∞, as a function of excitation

wavelength, λ. This data is usually presented as either the absorbance (log(1/R∞) or the

Kubelka-Munk, K-M function, i.e. F(R∞) = (1- R∞)2/2 R∞, which can be used 'as a proxy for the

typical absorption spectrum'27. DRS is often used to probe the absorption spectra of

semiconductor photocatalysts,27,28 and when combined with an action spectrum can help

identify the light-absorbing species3,29.

Note also that the action spectrum for an ideal photocatalytic system should be

independent of photo flux/irradiance. Thus, if the action spectrum for an ideal

photocatalytic system is recorded at several different average irradiance values, then the

resulting plots of η vs λ will superimpose on one another.

An ideal photocatalytic system is defined here as one in which: (i) the initial rate is

proportional to ρ at all λ and (ii) the test pollutant does not form a charge transfer complex,

CTC, nor does it absorb any of the wavelengths of excitation used in the study. A very nice

9

illustration of such an ideal system is provided by the work on Torimoto et al9, in their study

of the photocatalysed dehydrogenation of methanol (MeOH; 50 vol%) in water using

photoplatinized (2 wt%) anatase TiO2 (10 mg cm-3) from Merck (BET surface area: 13 m2 g-1).

In this case, the photocatalytic reaction can be summarized as:

Pt/TiO2 CH3OH → HCHO + H2 (6) λ

The rate of hydrogen evolution was measured as a function of excitation wavelength, using

a Xe/MC irradiation system (FWHM ca. 17 nm at all λ) at three different average irradiances

(0.1, 1.4 and 7.5 mW cm-2)9. The resulting kinetic data were used to generate the plot of

(normalized) photonic efficiency vs λ illustrated in figure 1. As expected for an ideal

photocatalytic system, the three action spectra for reaction (6), corresponding to the three

different average irradiances (i.e. 0.1, 1.4 and 7.5 mW cm-2), superimpose on one another

and give a very approximate fit to the DRS of the TiO2 powder (broken line in figure 1). The

ca. 15 nm red shift in the action spectrum compared to the DRS is attributed by the

authors10 to the relative large FWHM of the monochromatic light used in this work, when

compared to the high resolution of the DRS.

10

Figure 1: Plot of normalized (at 350 nm) photonic efficiency, η, for reaction (6), [MeOH] =50 vol%; [TiO2] = 0.010 g cm3, as a function of excitation wavelength, λ, measured using the following three different average irradiances: 0.1 (), 1.4(∆) and 7.5 () mW cm-2. The broken line is the normalised (at 350 nm) Kubelka-Munk data from the DRS of the Pt/TiO2 semiconductor photocatalyst powder. A Xe lamp/monochromator, i.e. Xe/MC, with FWHM = 17 nm was used as the light source. Data taken from reference 9.

Although the fit of the DRS to the action spectra illustrated in figure 1 is not very good, it

does indicate that the semiconductor photocatalyst is most likely the light absorbing species

in this system and, therefore, the reaction is probably a true example of photocatalysis. In

contrast, most photocatalytic systems, such as many of those listed in table 1, are non-ideal.

6. Simulation and testing ideal action spectra

Recently, the optical properties of dispersions of TiO2, from several different suppliers, have

been characterised30,31 in sufficient detail that a value for f can be calculated at a number of

different wavelengths for a particular dispersion of a relevant sample of TiO2 using the

expression:

f = 1 – 10Abs(λ) (7)

11

Where Abs(λ) is the effective absorbance of the TiO2 dispersion at wavelength λ, and given

by:

Abs(λ) = 0.434.κ(λ).[TiO2].b (8)

Where κ(λ), units: cm2g-1, is the specific absorption coefficient at λ, [TiO2] is the

concentration of the TiO2 dispersion (units: g cm-3) and b, units: cm, is the penetration depth

of the excitation light of wavelength λ, assuming its less than the pathlength of the reaction

cell. A value for b at each wavelength, λ, can be estimated by assuming that at b, 99% of

the light has either been lost by scattering or absorption, i.e. at b, Abs(λ) = 2, thus:

b = 2/(0.434.β(λ).[TiO2]) (9)

where, β is the Extinction (i.e. absorption plus scattering) coefficient. This calculation is

appropriate only if the photoreactor has an optical path length that is > b; which – as we

shall see - is likely in all cases considered here. The variations of β and κ as a function of λ

for P25 TiO2, as reported by Satuf et al.31, are illustrated in figure 2(a). The variation of b as

a function of λ can be calculated using this data, for example by assuming [TiO2] = 0.010 g

cm-3 and using eqn (8), so as to yield the plot of b vs λ in illustrated in figure 2(b). These

results show that, for [TiO2] = 0.010 g cm-3, b varies from ca. 0.007 to 0.011 cm as λ

increases from 295 to 405 nm; this being the case, it is reasonable to assume that in most

photocatalytic studies in which [P25 TiO2] = 0.010 g cm-3, the photoreactor's optical

pathlength is > b at all λ, i.e. all light is either scattered (and lost), or absorbed within the

photoreactor, and none is lost via transmission through the photoreactor.

12

Figure 2: Plots, as a function of λ, of (a) reported values of β and κ, and calculated values of b) penetration depth, b, and (c) fraction of light absorbed, f, by a 0.010 g cm3 dispersion of P25 TiO2. The red broken line in (c) corresponds to the action spectrum (η vs λ) for Pt/TiO2 (P25) for reaction (6), normalized here to provide the best fit, reported by Torimoto et al13.

13

Given the calculated variation of b vs λ, for [P25 TiO2] = 0.010 g cm-3, illustrated in figure

2(b), it is then possible to calculate how f varies as a function of λ, using eqns (8) and (7),

respectively, and the results of this work are illustrated in figure 2(c).

As noted above, in an ideal photocatalytic system the action spectrum should have the

same spectral profile as f vs λ, which in turn should be similar in shape to the DRS of the

semiconductor photocatalyst dispersion. Encouragingly, the measured action spectrum for

reaction (6) using platinized P25 TiO2, using [TiO2] = 0.010 g cm-3; [MeOH] = 50 vol%; FWHM

= ca. 23 nm, reported by Torimoto et al13, red broken line in figure 2(c), provides an

excellent fit to that of f vs λ predicted using the spectral data of Satuf et al. also illustrated in

figure 2(c).

In practice, a feature of an 'ideal' photocatalytic system appears to be one in which at least

one part of the overall photo-induced redox reaction, i.e. the reduction or oxidation, is

facile. For example, in reaction (6), the sacrificial electron donor, MeOH, is in vast excess

(50 vol% ≡12.4 M) and easily oxidized so that its reaction with the photogenerated holes on

a semiconductor such as TiO2, would be expected to be very fast.

As noted earlier, an important characteristic of an ideal photocatalytic reaction is that the

initial rate is proportional to ρ, or E, even at high levels, at all λ used to generate the action

spectrum. Although it is impractical to check the latter feature at all λ, a sample of

wavelengths should be tested. For example, in a study of reaction (6), Torimoto et al9

reported the initial rate to be largely independent of E over the range ca (0.05-0.1) – (5-10)

mW cm-2, when probed using 290, 350 and 380 nm light.

For all photocatalytic systems, even for an ideal one, as ρ is increased, a photon flux level

will eventually be reached, the threshold flux, ρ(threshold), much above which the initial

rate is no longer proportional to ρ, but rather ρθ, where θ is < 1 and typically tending

towards 0.532. Note that ρ(threshold) is used here as a mathematical construct, to help

identify the regions: (i) ρ << ρ(threshold), over which rate is proportional to ρ and (ii) ρ >>

ρ(threshold), where rate is proportional to ρ1/2.

In practice, this feature has been demonstrated in many different photocatalytic

systems9,33,34. A nice illustration of the ρ(threshold) region is provided by the work of by

Egerton and King33, in their study of the photocatalytic oxidation of 2-propanol to acetone,

14

by rutile TiO2, in which the initial rate was determined as a function of ρ over an

impressively wide range (ca. (0.002 to 103)x1016 photons cm-2 s-1) using a 250 W medium

pressure Hg lamp (unfiltered) as the light source A plot of these results is illustrated in

figure 3, along with broken red lines to highlight the regions in which the initial rate is

proportional to ρ (for low values of ρ) and proportional to ρ0.5. In this system ρ(threshold) is

ca. 4.7x1015 photons s-1, and, if we assume all photons are 365 nm, and an irradiation area =

2 cm2, this translates to an irradiance of ca. 1.3 mW cm-2, which would be considered a

moderate irradiance in most photocatalytic studies. The latter point is made to highlight the

fact that it would be wrong to assume for any photocatalytic system the values of

ρ(threshold) and E(threshold) are always exceptionally high, i.e. >> 10 mW cm-2 at 365 nm

say; whereas, in practice the threshold irradiance can be < 1.5 mW cm-2.

Figure 3: Plot of log(ri) vs log(ρ), reported by Egerton and King33 in their study of the photocatalytic reaction (1), using rutile TiO2 and P = 2-propanol. The broken lines are the lines of best fit when the initial rate is initially (at low ρ) proportional to ρ and (at higher ρ) then proportional to ρ0.5. The threshold photon flux for this system appears to be ca. 4.7x1015 photons s-1 (i.e. ca. 1.3 mW cm-2, assuming all photons are 365 nm and an irradiation area of ca. 2 cm2; Hg (med. pressure lamp).

15

The often made argument32 for the change of dependence of initial rate upon ρ is that at

low ρ the overall photocatalytic reaction is dependent directly upon the rate of generation

of electron-hole pairs and, therefore, ρ, because direct electron-hole pair recombination is a

minor process, but at high irradiances, electron-hole pair recombination dominates and the

overall rate tends to be dependent upon ρ0.5. In practice10,35, at the moderate irradiances

commonly employed in semiconductor photocatalysis (ca. 0.5-3 mW cm-2) it is not

uncommon for the rate to be depend directly upon ρθ, where 0.5 < θ < 1.

So far we have considered the features of an ideal action spectrum, i.e. one in which θ = 1

and so eqn (5) is obeyed at all λ Any deviation from this ideal spectral profile can be

considered a distortion (from the ideal), the cause of which may be due to one, or more

effects which include: (i) variation in emission irradiance/intensity, (ii) crystal phase and (iii)

pollutant, absorption; and, each of these effects is discussed below.

7. Intensity distorted action spectra

The intensity distortion effect is readily demonstrated by considering what the action spectrum of a simple photocatalytic system would look like if the initial rate is proportional to ρθ rather than ρ, where 0.5 <θ < 1.

It follows from eqn (5) that, at any excitation wavelength, the measured photonic efficiency,

η, will be given by the expression:

η(θ) = kθ(αfρ)θ/ρ (10)

Where kθ is a proportionality constant (units: moles/s) which depends upon: (i) the initial

concentration of P: [P]o and (ii) the intrinsic activity of the photocatalyst, i.e. the probability

that an absorbed ultra-bandgap photon leads to a reaction; α is a proportionality constant

of convenience, with a value of unity, but with units: s einsteins-1, so that it renders the

collection of terms, αfρ, unitless and simplifies the units of kθ and any subsequent

mathematical manipulations of eqn (10). If we assume that at ρ(threshold), η(θ =1) = η(θ =

0.5, it follows that:

k0.5/k1 = (αfρ(threshold))0.5 (11)

so that

16

η(θ=0.5)/η(θ=1) = {ρ(threshold)/ρ}0.5 (12)

And so in all cases where ρ > ρ(threshold), then the greater the value of ρ, compared to

ρ(threshold), the lower the value of η, i.e. when θ = 0.5 (and, in fact, more generally when θ

< 1). Thus, when θ < 1, η will be dependent upon ρ, at any excitation wavelength, which is

in striking contrast to an ideal system, where θ = 1 and η is independent upon ρ As a result,

for any photocatalytic system under study in which θ < 1, the resulting action spectrum will

appear distorted from that of the true, i.e. 'ideal', action spectrum (for which θ = 1).

In order to gain an insight into the extent of this distortion effect, consider a TiO2

photocatalytic system in which the fraction of light absorbed, f, as a function of λabsorbed is

as illustrated in figure 2. If the photon flux used was sufficiently low that it never exceeded

ρ(threshold), it follows from eqn (5) that η(θ = 1) = k1f and the action spectrum would have

the same spectral profile as that in figure 2(c), as illustrated in figure 4(a) – solid red line,

assuming k1 = 1; i.e. the system would generate an ideal action spectrum.

Now let us consider the case where the photon flux used was sufficiently high that at all λ, it

always exceeded ρ(threshold). The combination of eqns (5) and (12) yields the following

relationship between η(θ = 0.5) and ρ:

η(θ=0.5) = k1f.{ρ(threshold)/ρ}0.5 (13)

Note: this expression only holds provided ρ > ρ(threshold). In order to make these

calculations more pertinent, we require a typical lamp profile and figure 4(b) illustrates

some extreme lamp profiles, namely, those for: (i) a lamp with a photon flux density (i.e.

ρ/A, where A = irradiation area, that is independent of λ (set at 1014 photons nm-1 cm-2 (≡

0.054 mW nm-1 cm-2 at 365 nm), and the typical emission spectra of (ii) a Xe and (ii) Xe/Hg

1000W lamp. For simplicity we shall set ρ(threshold) to be ca. 3.5x1013 photon nm-1 cm-2 (≡

0.019 mW nm-1 cm-2 at 365 nm), so that at all λ, ρ > ρ(threshold), and eqn (13) can be used

to calculate the resulting action spectra, which are illustrated in figure 4(a). These

calculated action spectra show that only if ρ is constant at all λ will the action spectrum

(broken line in figure 4(a), have the same spectral profile as that of the ideal action

spectrum, i.e. the action spectrum of same system, but carried out at ρ < ρ(threshold).

17

In contrast, in the two other cases, using emission profiles for real lamps, the calculated

action spectra are distorted versions of the ideal action spectrum, and the distortion is due

to the emission profile of the lamp. The most dramatic of these distortions is the action

spectrum associated Xe/Hg lamp, i.e. the blue spikey line in figure 4(a), which highlights the

striking effect a markedly variable lamp emission spectrum has on an the action spectrum

for a photocatalytic system in which rate is proportional to ρθ, where 0.5 < θ < 1; in this

example θ = 0.5. A quick comparison between the action spectrum one might expect for a

Xe/Hg lamp and the emission of that lamp (illustrated in figures 4(a) and 4(b), respectively)

reveals that the action spectrum is a composite of the f vs λ spectrum of the semiconductor

photocatalyst and the 'negative' of the lamp emission spectra, exhibiting peaks and troughs

in the action spectrum where there are troughs and peaks in the emission spectrum.

Clearly, this simulation suggests that it would be very unwise to run an action spectrum

using a Xe/Hg lamp, since it is very possible that at the emission peaks in the spectrum the

condition ρ > ρ(threshold) will be met, so that θ < 1 and the resulting action spectrum will

exhibit a series of peaks and troughs.

18

Figure 4: (a) Action spectra for a P25 dispersion calculated using eqn(13) and the variation of f vs λ illustrated in figure 2(c) combined with the emission spectra illustrated in (b). The solid red line corresponds to the action spectrum of the system which will be observed using any light source, provided ρ < ρ(threshold), i.e. θ = 1 at all λ. The other action spectra have been calculated assuming ρ > ρ(threshold) and θ = 0.5 at all λ, using either: (i) a constant ρ at all λ light source (broken black line), (ii) a Xe lamp (solid black line) and (iii)a Xe/Hg lamp (a solid blue line).

19

In fact, the mistaken use of a Xe/Hg lamp to record an action spectrum appears to be at the

heart of the claim of spectral sensitivity reported by Emeline et al8. in their study of the

photocatalysed oxidation of 4-chlorophenol by P25 TiO2. i.e.

P25 TiO2 4-CP + O2 → oxidation products (14) UV

Thus, Emeline et al.8 reported the action spectrum illustrated in figure 5(a) for reaction (14),

using [4-CP]o = 0.2 mM, [TiO2] = 0.3 g dm-3 and pH = 3 and a 1000W Xe/Hg lamp coupled to a

monochromator (FWHM = 10 nm). A brief inspection of this η vs λ(ex) action spectrum

reveals that it comprises a series of peaks and troughs (see figure 5(a), black line) which, the

authors claim8, demonstrates that the photocatalysed oxidation of 4-CP by oxygen,

photosensitised by P25, i.e. reaction (1), under the conditions described above, is spectrally

dependent. Note that such spectral sensitivity is counter to the classical band model of

semiconductor photocatalysis, in which the photogenerated charge carriers are thermalized

before reacting8. However, when the photon fluxes used to calculate the different values of

the various wavelengths are superimposed onto the action spectrum, as illustrated in figure

5(a), it is obvious that the troughs in the action spectra occur at the peaks in the emission

spectrum and that the reported action spectrum for reaction (14) illustrated in figure 5(a) is

the 'negative' of the sampled emission spectrum of the Xe/Hg lamp, with FWHM = 10 nm.

This feature is as predicted above, assuming that, at the peaks in the emission spectrum at

least, the initial rate is not proportional to ρ. This finding clearly calls into question any

claim of spectral sensitivity for this system, or any system studied by any group using a

Xe/Hg lamp in which the peaks and troughs in the action spectrum are matched by the

troughs and peaks in the emission spectrum of the excitation lamp.

It should be noted at this point that in the same paper8 the authors reported that at 334,

365 and 380 nm, they found, for this system, that the initial rate of reaction (13) was

proportional to ρ; which – if true – would rule out the suggestion that the peaks and troughs

in the action spectrum were due θ being < 1 at the Xe/Hg lamps' emission peak wavelengths

at least, if not at the other λ. However, the irradiance at 365 nm at least (ca. 12.3 mW cm-2)

is very high and others15,35,36, studying the same system have reported θ < 1, with typically θ

= 0.5, at much less levels (2-6.4 mW cm-2).

20

Figure 5: Action spectrum (black line, data points) for reaction (1), [P25 TiO2] (0.3 mg cm-

3), [4-CP]o = 0.2 mM and pH (pH 3), (a) as reported by Emeline et al.8 , generated using the photon fluxes illustrated by the blue, broken line and data points for a 1000W Xe/Hg lamp/MC (FWHM = 10 nm) and (b) as reported by this group using the irradiances illustrated by the blue, broken line and data points for a 1000W Xe lamp. The red line corresponds to the diffuse reflectance absorbance spectrum of the photocatalytic reaction solution.

21

In order to test the claim of spectral sensitivity for this system the Mills group36 recently

reported the action spectrum for reaction (14) under identical conditions to those reported

by Emeline et al8, but using a 1000W Xe lamp, and the results of this work are illustrated in

figure 5(b). A brief inspection of the action spectrum for the 4-CP/P25 TiO2/O2 system,

recorded using a Xe lamp/MC as the excitation light source illustrated in figure 5(b) reveals

no evidence of spectral sensitivity, i.e. no peaks and troughs, but a striking similarity to the

diffuse reflectance spectrum (absorbance) spectrum of the P25 dispersion in the

photocatalytic system, also illustrated in figure 5(b) (red line). These results provide strong

support for the proposition that the peaks and troughs in the action spectrum of the 4-

CP/P25 TiO2/O2 system, recorded using a Xe/Hg lamp/monochromator are due to the lamp

profile, i.e. there are a lamp artefact, and so do not provide evidence for spectral sensitivity

in this system36.

The distortion of an action spectrum due to the emission spectral profile of the lamp,

provided that θ < 1 at one or more wavelengths, is most apparent using a lamps with very

spikey emission spectra, such as a Xe/Hg or Hg (medium or high pressure) lamp, however,

researchers should be aware of the fact that even a Xe lamp can distort an 'ideal' action

spectrum somewhat, as illustrated in figure 4(b), although, not very dramatically. A possible

illustration of this distortion is the action spectrum reported by Torimoto et al.13 for the

photocatalytic oxidation of acetic acid (AA), i.e.

TiO2 AA + 2O2 → 2CO2 + 2H2O (15) UV

In this work, amongst other commercial forms, P25 TiO2 was used with [TiO2] = 10 mg cm-3

and [AA] = 5 vol%. Previous work by the same group9, using Merck TiO2, had established at

sample wavelengths: 290, 350 and 380 nm, that the initial rate was proportional to ρ0.5, over

the irradiance range: 0.1 – 5 mW cm-2, and so it will be assumed θ = 0.5 in this same system

using P25 as the photocatalyst. In their subsequent study13 of the action spectrum for P25

the intensities used were: 1.4, 2.1, 2.7 and 3.6 mW at 350, 370, 385 and 410 nm,

respectively, using a Xe lamp/MC system with a FWHM of ca. 23 nm. The normalized action

spectrum reported for this system is illustrated in figure 6, which matches quite well that of

the normalized (at 350 nm) simulated action spectrum, (broken red line in figure 6) once the

distortion by the Xe lamp emission spectrum has been included. The later simulated

22

spectrum was generated using the spectral data of Satuf et al.31 for P25 TiO2,see figure 2(a),

and the same method as used to generate the action spectra in figure 4. For comparison,

the solid red line in figure 5 is the predicted shape of the action spectrum for the same

system with no such distortion, i.e. if θ = 1, so that the action spectrum is independent of

the emission spectrum of the excitation light. A comparison between the solid and broken

red lines reveals the predicted degree of distortion the action spectrum undergoes when θ =

0.5 and a Xe lamp/MC (for which ρ is not constant at all λs) is used as the excitation source

and P25 TiO2 is the photocatalyst. From this it is clear that the distortion is not very great

from 400-350 nm at least, since over this range the Xe lamp emission spectrum is fairly flat,

see figure 4(b). This would suggest that the Xe lamp is the best light source for recording

the action spectrum of UV-absorbing photocatalysts, even if efforts are not made to make ρ

the same at all wavelengths when recording the action spectrum.

Figure 6: Action spectrum plot, normalized at 350 nm, for reaction (14), [AA] =5 vol%; [P25 TiO2] = 0.010 g cm3, using a Xe lamp/MC with FWHM = 23 nm13. The broken red line is the normalized (at 350 nm) action spectrum predicted using the same process as used in figure 2, and assuming distortion by the Xe lamp spectral profile because θ = 0.5. The solid red line is that predicted assuming no distortion, since θ = 1.

23

Mixed crystalline phase distorted action spectra

In semiconductor photocatalysis the most important two crystalline phases of any

semiconductor material tested to date are the anatase and rutile forms of TiO2, with

bandgaps 3.2 and 3.0 eV, respectively. This difference in bandgap is reflected in a difference

in the conduction band potentials of: -0.20 and 0.01 V vs NHE at pH 0, for anatase and rutile

TiO2 respectively37. Evidence has been found that these two different phases exhibit

different activities depending upon the photocatalytic reaction that is under test. For

example, it has been found13 that in the photo-oxidation of organic compounds, such as

acetic acid, AA, in reaction (15), the anatase crystalline phase of TiO2 is more reactive than

rutile, whereas in the photocatalysed reduction of silver nitrate by water, i.e.

TiO2 4Ag+ + 2H2O → 4Ag + O2 + 4H= (16) UV

it appears that rutile is the more active phase. This should not affect the shape of the action

spectra recorded using any pure phase, such as pure anatase or pure rutile TiO2. However,

it can lead to striking differences when a mixed phase, as a P25 TiO2 is used (80:20

anatase:rutile38) and this has been very well illustrated by the work of Torimoto et al.13 in

their study of the action spectra of a wide variety of pure and mixed forms of TiO2 when

used to promote reactions (6), (15) and (16). Figure 7 illustrates their findings for P25 TiO2,

which shows that the action spectrum for the photodeposition of Ag from AgNO3 is what

might be expected for the narrow bandgap rutile form of TiO2, whereas that for the

oxidation of AA corresponds to that for the larger bandgap crystalline phase of this

semiconductor, anatase. Interestingly, from the profile in figure 7, and as noted by the

authors, it appears that when platinized P25 is used to photocatalyse reaction (6), the

dehydrogenation of MeOH, neither phase is dominant, which helps justify its earlier use

here as an example of an ideal photocatalytic system, see figure 1. The results in figure 7

show that in a mixed crystalline phase oxide such as P25 TiO2, the shape of the action

spectrum may depend on the selectivities of the two crystalline phases for the reaction

under study and, under such circumstances the overall action spectrum may appear a

distortion form the ideal, in which both phases are equally active.

24

Figure7: Reported normalised action spectra for reactions: (6), open circles red broken line; [MeOH] = 50 vol%, (13), ; [AA] = 5 vol%, and (14), ; [AgNO3] = 25 mM. In all cases: [TiO2] = 10 mg cm-3 and irradiation source: Xe lamp/MC; FWHM = 23 nm13.

Pollutant distorted action spectra: Dyes

The process of dye photosensitization is probably the most well-known and established of

the processes associated with pollutant distorted action spectra. A classic example of this

process is the N-deethylation of Rhodamine B (RhB) adsorbed onto CdS in aerated aqueous

solution reported by Watanabe et al. in 197739. This process is readily monitored by UV/Vis

absorption spectroscopy since it is associated with a hypsochromic shift in the absorption

maximum for the dye from 555 nm, see figure 8(b), to 498 nm. A typical action spectrum

reported for this system is illustrated in figure 8(a), where [RhB] = 1 mM and [CdS] = 20 mg

cm-3, and reveals significant activity at excitation wavelength > 540 nm, above which the CdS

does not absorb, although it does scatter, as illustrated by its reported absorption spectrum

illustrated in figure 8(b) (broken and dotted line). The action spectrum for the RhB/CdS

system has a peak at ca. 600 nm, which is not too dissimilar to that of RhB adsorbed onto

CdS (λ(max) = 577 nm), since the former forms multimers when adsorbed on the powder,

25

Figure 8: (a) Action spectra for dye de-alkylation: (i) CdS (20 mg cm-3) /RhB (1 mM)/O2 (air sat.), solid line; 39, and (ii) CdS (6 mg cm-3) /MB (0.5 mM)/O2 (air sat.),broken line, 25; (b) absorption spectra of RhB (solid line), MB (broken line) adsorbed on CdS (measured via the reflectance spectrum of the dyed powder and CdS powder alone (solid line); Xe lamp/MC, FWHM = ca. 20 nm.

26

see figure 8(b). This action spectrum is interesting in that at wavelengths < 540 nm, the

value of η is approximately constant and equal to 0.03; further work shows that the latter

value is independent of [RhB]. This feature is consistent with the suggestion of the

authors39 that at λ < 540 nm the deethylation of RhB is photocatalytic, with η = ca. 0.03,

whereas above 540 nm it is a dye-sensitised process, with η = ca. 0.46 when [CdS] = 200 mg

cm-3 These same group also studied25 the demethylation of Methylene Blue (MB) by CdS

and found that the action spectrum, illustrated in figure 8(a) (broken line, open circles)

exhibited no evidence of a dye–sensitisation process, because the electronically excited

state of MB is not sufficiently reducing to inject an electron into the conductance band of

the CdS. As a result, the recorded action spectrum resembles that of the absorption

spectrum of CdS alone, illustrated in figure 8(b), thereby indicating that the demethylation

of MB reaction is, in striking contrast to RhB, solely a photocatalytic process25. The

absorption spectrum of MB adsorbed on the CdS, illustrated in figure 8(b), reveals that it

too, like RhB, forms multimeric species when adsorbed on CdS powder (λ(max) = 612 nm).

As illustrated by the data in figure 8(b), the CdS/RhB and CdS/MB systems are ones in which

the adsorbed dye absorption spectrum overlaps with that of the semiconductor and so, as a

consequence, it is essential to record an action spectrum in order to identify whether the

semiconductor (in the case of photocatalysis), or dye (in the case of dye sensitisation or

photolysis), is responsible for the observed photo-induced spectral changes in the

absorption spectra of the reaction solutions. However, in cases where the dye absorption

spectrum clearly doesn't overlap with that of the photocatalyst, as we shall see below, then

it may be unnecessary to run an action spectrum to substantiate the photocatalytic nature

of the reaction under test, provided the excitation light used is of a wavelength which the

photocatalysts absorbs, but the absorbed dye does not.

It is clear from the UV/Vis absorption spectra that of RhB and MB, and probably most other

dyes, illustrated in figure 8(b), that when they are absorbed strongly onto the surface of the

photocatalyst, the normally narrow visible bands of the monomeric forms of the dyes

(typically FWHM = 40 nm), are significantly broadened, so much so that overlap with the

absorption spectrum of the photocatalyst is possible. This broadening can render the

system inappropriate for the unambiguous assessment of the photocatalytic activity of the

semiconductor in the system. However, such significant absorption often depends both on

27

the dye concentration and, more importantly, the pH of the system40-42. For example most

cationic dyes, such as MB, do not absorb on TiO2 when the pH is much more acidic (e.g. pH

2) than the pzc of TiO2 (pH = 6.6), since the surface of the titania is then very positively

charged61,63 under such conditions; whereas, in contrast, in alkali (pH 11) MB is very strongly

absorbed onto TiO2. Similarly, the anionic dye, AO741,42, does not absorb strongly on TiO2 at

pH 11, but does at pH 2. Thus, if a dye is to be used to test the photocatalytic activity of a

new material, and if an action spectrum is not run, because the dye used has an absorption

spectrum that appears very well separated from that of the photocatalyst, then it is

important to note that the absorption spectrum that needs to be considered is that for the

dye adsorbed on the photocatalyst under the conditions of the experiment and not the dye

alone in aqueous solution. The latter is best determined by recording the diffuse

reflectance spectrum of the reaction solution, rather than that of a dried form of the

powder plus dye.

In spite of the significant potential pitfalls of using dyes7 to probe the photo-oxidation

capability of a photocatalyst, via reaction (1), the use of dyes, especially MB and acid orange

7, AO7, as the test pollutant continues to be a very popular approach to the assessment of

the photocatalytic activity of any new photocatalyst, via the photo-oxidised mineralization

of the dye by dissolved oxygen. Its popularity lies in the striking nature of the reaction

(highly colored – to colourless) and the ease with which it can be performed, as it requires

only a UV/Vis spectrophotometer. However, in such work, unless an action spectrum of the

system is run7, in order to demonstrate that it is exclusively a photocatalytic process, there

will always be a doubt as to whether the observed dye bleaching process was due in part, or

wholly, to dye-sensitisation via reaction (2), or dye photolysis. This confirmatory step is

particularly important when the novelty of the claim is visible light photocatalysis for

systems in which the only test pollutant tested was a dye, usually MB. Obviously, in such

cases, in the absence of an action spectrum, the claim of visible light photocatalytic activity

should be considered ambiguous at the very least. Indeed, there have been calls5 for the

use of dyes as test pollutants in monitoring photocatalytic reactions to be abandoned,

unless both the action spectrum and absorption spectrum have both been run for the

system under study, to prove, as in the case of the demethylation of MB by CdS25, that the

reaction is truly photocatalytic (see figure 8).

28

One possible way in which a dye may be used to test photocatalytic activity and avoid the

accompanying concern regarding dye photosensitization is to study the photocatalysed

reduction of a dye, such as resazurin, Rz (blue coloured; λ(max) = 605 nm), by a sacrificial

electron acceptor, such as glycerol, in anaerobic solution, i.e.

SC Glycerol + Rz → glyceraldehyde + Rf (17) hν ≥ Ebg

where Rf is resorufin (pink coloured; λ(max) = 585 nm)43. This is in striking contrast to the

usual use of dyes in photocatalysis in which the photo-oxidation of the dye is studied in the

presence of a sacrificial electron acceptor, such as O2, as in reaction (2). In the absence of

O2, as in reaction (17), the photo-bleaching of the dye via the traditional dye sensitisation

route, i.e. reaction (2), is not possible. As a consequence, provided the excited state of the

dye is not quenched by glycerol, as in the case of Rz, then reaction (17) can only occur via a

photocatalytic process44. A simple demonstration of this was reported recently by this

group using a variety of different semiconductors, including CdS6. The photocatalysed

reduction of Rz to Rz was monitored spectrophotometrically, and a typical set of results for

CdS are illustrated in figure 9(a), which reveal that the blue to pink colour change is rapid

when ultra-bandgap light is used(λ(excit) = 455 nm; 20 mW cm-2). A brief inspection of the

absorption spectra of the Rz, λ(max) = 608 nm, and Rf, λ(max) = 585 nm, see figure 9(a),

reveals little or no absorption at 455 nm by either Rz or Rf, but significant absorption by CdS,

see figure 8(b). When the CdS/Rz system is irradiated with light which Rz does absorb and

CdS doesn't, e.g. λ(excit) = 617 nm; 15 mW cm-2, the observed photobleaching is negligible

over the 5 min time period required to complete reaction (17), using 455 nm light (see

figures 9(a) and 9(b)6. However, prolonged irradiation at 617 nm of the CdS/Rz system does

reveal that a very slow dye photobleaching process occurs, but at a rate that is unchanged in

the absence of CdS and/or glycerol, thereby indicating that it is due to dye photolysis. These

finding suggest that a dye, like Rz, could be used to assess the photocatalytic activity of UV

and visible light absorbing semiconductors, via a reductive photocatalytic reaction, such as

reaction (17), using 365 or 455 nm radiation, provided the photocatalysed reaction was

much faster than the slow photolysis of the dye.

29

Figure 9: (a) Recorded spectral changes for the photocatalytic reaction (16), mediated by CdS in aqueous solution ([CdS] = 0.5 wt%; [glycerol] = 10 wt %)6, carried out in a 1 cm cuvette, using a 455 nm LED (20 mW cm-2). The spectra were recorded every 20 s; (b) Recorded absorbance at 608 nm (λ(max) Rz) as a function of irradiation time, t, for the same solution in (a) when irradiated with 455 nm light (20 mW cm-2), solid line, or 617 nm light (15 mW cm-2), broken line.

30

However, the most obvious way to avoid concerns of dye-sensitisation is to use simple

organic test pollutants that will not absorb significantly the excitation light, such as MeOH,

via reaction (6), although, in this case, this will require the semiconductor to be platinized

beforehand, since photocatalysts most have high overpotentials for water reduction17.

Alternatively, acetic acid, AA, via reaction (15), can be used as the test pollutant, as has

been demonstrated by Nishijima and co-workers44 in a study of the photocatalytic activity of

a sulfur-doped TiO2 visible light absorbing photocatalyst, the action spectrum for which, and

that of an undoped, UV-only absorbing sample of P25 TiO2 for comparison, are illustrated in

figure 10.

Figure 10 Diffuse reflectance and action spectra for the photocatalysed oxidation of AA by dissolved O2, i.e. reaction (14) using either P25 (, solid line) or S-doped TiO2 (, broken line) as the photocatalyst; [TiO2] = 10 mg cm-3, [AA] = 0.05 mM; Xe lamp/MC, ca. FWHM = 17 nm44.

The fits of the action spectra for reaction (15) to the DRS absorption spectra of the

semiconductor photocatalysts are reasonable, although not great, as illustrated in figure 10;

these less than perfect fits are attributed by the authors44 to the large FWHM (17 nm) used

to generate the action spectrum. However, it is clear from the action spectra illustrated in

figure 10 that the S-doped photocatalyst action spectrum does show clear evidence of

31

visible light photoactivity, and that this feature is absent from that of P25 TIO2. However,

even with this apparently simple system, using TiO2 alone, the reaction mechanism is is

complex, since below an irradiance of ca. 5 mW cm2 the rate depends upon ρ0.5, but

depends upon ca. ρ-1 above this threshold. These two features, the authors9 tentatively

suggest reflect the radical chain nature of mechanism, which probably involves peroxy

radicals.

Pollutant distorted action spectra: Charge Transfer Complexes (CTCs)

If the pollutant doesn't absorb visible light itself but instead forms a visible-light absorbing

(ligand to metal) charge transfer complex, i.e. CTC, with the semiconductor, then it is

possible that the visible light induced disappearance of the organic pollutant observed upon

irradiation may be due to electronic excitation of the CTC and not photocatalysis6. In the

former case, the electron is photoexcited directly from the ground state of the adsorbate to

the semiconductor, which is invariably TiO24. As in the dye sensitisation process, the

oxidised form of the adsorbate may be sufficiently unstable so as to degrade, although

complete oxidative mineralisation is not usually achieved via this process. Numerous

groups have been reported cases of organics forming CTC's with TiO2, including ones that

can be described as: phenolic (such as catechol, salicylic acid and phenol), hydroxyl (e.g.

cyclodextrin) and carboxylic (e.g. EDTA, oxalic, formic and citric acid and aromatic carboxylic

acids)4.

In many cases, the CTC is unstable when exposed to visible light and thus the latter

promotes the overall oxidative degradation of the original (non-complexed) organic with

irradiation time. Clearly, this photochemical feature could be easily misinterpreted as an

example of visible photocatalysis, whereas in fact it is a version of dye-sensitisation; where

the CTC is the light-absorbing species. These CT complexes are usually identified via the

appearance of additional absorption bands, often in the visible, in the DRS. As a

consequence, it appears that before any unambiguous claim of visible light photocatalysis

can be made, it is also essential to record the absorption spectrum, usually the DRS, of the

photocatalytic system in order to identify where the various components absorb, i.e. the

substrate and semiconductor, both alone and when combined to create the overall

photocatalytic system under investigation.

32

Probably the most well-known of the CTC's is that formed between TiO2 and 4-CP4. Choi

and his co-workers have reported that 4-CP weakly absorbs onto TiO2, most notably ST1

(Ishihara Co.) with a specific surface are ca. 340 m2 g-1, and as such can be degraded, to form

CO2 and Cl-, under visible light, and that the rate can be correlated with oxide specific

surface area4. These same workers also reported that the combination of 4-CP and P25 TiO2

form a weak CTC9, although others, looking at the same system report that such

complexation is negligible45. Choi et al.4 also note that: 'the LMCT induced degradation of

phenolic compounds under visible light is much slower than that of bandgap-excited

photocatalysis under UV-irradiation'.

An interesting example CTC photochemistry is that reported by Higashimoto et al46. in their

study of the photocatalysed oxidation of benzyl alcohol (BA) to benzylaldehyde by dissolved

oxygen using TiO2 in acetonitrile. This work is notable, not so much for the process, but

rather the form46 of the quantum yield, φ, versus λ, as illustrated in figure 11. As noted

earlier, φ vs. λ plots are not common in heterogenous photocatalysis , and a brief inspection

of the data reveals that the φ, vs λ plot matches quite well that of the absorption spectrum

of the BA/TiO2 CTC. However, theoretically this should not be the case, since for an ideal

photocatalytic system, by definition:

φ = η/f = k (18)

i.e. unlike η, φ should be independent of the fraction of light absorbed, f, and most likely a

constant at all wavelengths.

Any attempt to measure the φ vs. λ profile of a photocatalytic system requires details of

how the f vs λ spectrum was determined in the first place, since without which the values of

η cannot be converted to φ at each λ. In the case of Higashimoto et al.46 and their study of

the BA/TiO2 system, no such details are given, suggesting that the researchers did not

actually determined f vs λ for the BA/TiO2 system. This observation, along with the fact that

the action spectrum plot illustrated in figure 11 matches quite well with that of the

absorption spectrum of the system, strongly suggests that the plot of φ vs λ in figure 11, is in

fact a plot of η vs. λ, and that the researchers18 have mistakenly assumed that all the

incident light is absorbed by the system, i.e. none is lost via scattering or transmission, so

that f = 1 at all wavelengths, whereas in practice such losses are usually significant30,31.

33

Figure 11: UV/Visible absorption (solid line) and action spectrum, ,of the BA (5µmol cm-3)

/TiO2 (5 mg cm-3) system in MeCN (10 cm3). 500 W Xe lamp/MC; FWHM = 13.6 nm46. The

dotted line is the absorption spectrum of TiO2 alone.

Catalyst-distorted action spectra:

It is well known that deposits of noble metals, such as Pt, can greatly enhance the rate of a

photocatalytic process, most notably that of the reduction of water for which the

overpotential is so high for the semiconductor alone, that the rate is negligible in the

absence of such deposits47. Thus, the photocatalysed reduction of water, and concomitant

oxidation of a sacrificial electron donor, SED, such as MeOH, see reaction (6), requires a

platinized photocatalyst, or at least one with a noble metal deposit, since all have low

overpotentials for water reduction9. Similarly, the photo-oxidation of organics by oxygen,

i.e. reaction (1), is often enhanced by the presence of deposits of noble metals, such as Pt,

since this appears to facilitate the reduction of the oxygen by the photogenerated

conductance band electrons47,48.

34

In the case of Pt, nanodeposits on TiO2 usually produce a grey powder which exhibits an

action spectrum that is of the semiconductor alone, for example see figure 1, and which

shows no evidence of any unexpected visible light activity9. In contrast, recent work on

gold-modified titania (Au/TiO2), reveals evidence of visible light activity, which has been

attributed to photoexcitation of the localized surface plasmon resonance (LSPR) of the gold

particles, in a process not too dissimilar to dye-sensitisation, i.e. reaction (2)22,48,49. Thus,

the latter is an example of a case where catalyst deposits can distort the expected action

spectrum of a photocatalytic system, if they are LSPR photoactive, in much the same way a

dye can, via a sensitization process.

A very good example of the above is provided by the work of Kowalska et al.22, in their study

of the action spectrum for reaction (1), using Au (2 wt%)/TiO2 and P = 2-propanol, the

results of which are illustrated in figure 12. The match between the action spectrum and

the DRS of the Au/TiO2 (Aldrich-rutile) photocatalyst provides compelling evidence that over

this wavelength range the Au particles (ca. 80 nm) are excited via their LSPR and that most

likely this results in the injection of an electron from the Au particles to the TiO2, which is

then able to reduce adsorbed O2, whilst the electron-deficient Au oxidises the 2-propanol to

acetone22. Although, to date, most reported photocatalyst-related SPR processes are not

every efficient, it is worth noting their existence at least, especially when considering cases

of apparent visible light photocatalytic activity, in which nanoparticles of a metal are often

used to enhance the efficacy of the photocatalytic process.

35

Figure 12: Action spectrum for the oxidation of 2-propanol by Au/TiO2 (Aldrich: rutile), ,

and bare TiO2, , and DRS (thick solid line) of the Au/TiO222.

General guide to running an action spectrum

An action spectrum is a plot of η vs λ which, in an ideal classical photocatalytic system,

should have the same shape as the plot of f vs λ, which, in turn, should be similar in shape

to the DRS/absorption spectrum of the system. Obviously, for any photocatalytic

measurement the system should be as well-defined as possible, so that others might be able

to reproduce it. Thus, all the usual experimental parameters should be defined, such as

nature and source of the reactant(s), their concentrations, the nature and source of the

semiconductor photocatalyst and its concentration, pH, temperature, ionic strength,

solvent, light source, irradiation set up, photoreactor design and construction, method and

degree of agitation (especially for powder dispersions), nature and method of gas (usually

air) saturation. However, more particularly, when recording an action spectrum efforts

should be made to:

36

(1) Use a light source devoid of spikes in its emission spectrum; e.g. use a Xe lamp but never

a Hg/Xe or Hg lamp

(2) Use a monochromator with (1) to create light at each wavelength with a low FHMW,

preferably ca. 10 nm

(3) Ensure that the %variation of ρ across the wavelength range under investigation is as

small as possible, typically < 30%, but, ideally, 0%

(4) Before running the action spectrum, determine the variation in the initial rate of the

light-induced reaction as a function of ρ at 3 different wavelengths that span the active

wavelength range under investigation, so as to determine if it is reasonable to assume θ = 1,

i.e. that the system is 'ideal' and will not be prone to intensity distortion effects.

(5) Before running the action spectrum, record the DRS/absorption spectrum of the

individual components of the photocatalytic system, i.e. the semiconductor and substrate,

as well as the photosystem as a whole. Study these spectra for evidence of: (i) substrate

aggregation/adsorption, Pads, on the semiconductor photocatalyst and/or (ii) CTC formation.

Evidence of (i) and/or (ii) will undermine any claim of pure photocatalytic activity. Any

significant overlap of the absorbance due to P, Pads, or CTC with that of the semiconductor

photocatalyst alone will undermine any claim that the process under study is purely

photocatalytic.

(6) Determine the action spectrum of the system based on initial rates and using equation

(4) and compare the spectral profile of the action spectrum with that of the DRS/absorption

of the photocatalytic system in order to identify the light absorbing species and help

support any claim that the reaction is purely photocatalytic.

Note that even for a well-defined photocatalytic system, reproduction of the exact values of η may

prove difficult from laboratory to laboratory, since the absolute value of the rate depends upon so

many different parameters, including the light flux in the photoreactor, which depends amongst

other things, the degree of scattering and reflection of the incident light, the beam size and shape

and the reactor design . As a consequence, some have suggested that individual values of η 'have

little, if any, meaning'50. Thus, in recording the action spectrum, i.e. η vs. λ, of a photocatalytic

system, it is not the absolute values of η that are of interest per se, but rather the spectral profile

they reveal, which shows whether the system is photocatalytic or not.

37

Conclusions

The recording of an action spectrum, as well as the absorption or DRS spectrum, of a

photocatalytic system is an increasingly important part of any study of photocatalytic

systems, especially ones in which it is claimed that the photocatalyst is visible light

absorbing. The ideal action spectrum of a classical photocatalytic system will have the same

spectral profile as the fraction of light absorbed by the semiconductor as a function of λ,

with an initial rate that is proportional to ρ at all wavelengths; unfortunately, most

photocatalytic systems exhibit non-ideal action spectra. There are several major possible

causes for non-ideal action spectra, including: light intensity effects, crystal phase effects,

dye-sensitisation photolysis, CTC formation and LSPR absorption by a deposited noble metal

catalyst. Such non-ideal behavior could lead to mistaken claims of visible light

photocatalysis. Indeed, it has been suggested3 that it is unwise to use dyes to probe the

activity of a photocatalyst, especially visible-light absorbing photocatalysts. In the latter

case the photocatalyst is used to promote the oxidation of the dye by O2, or another

sacrificial electron acceptor. Recent results indicate6 that dyes may be appropriate for

assessing photocatalytic activity when they are reduced, and a sacrificial electron donor is

oxidized, i.e. via reductive photocatalysis. Whatever the assessment method, the recording

of an action spectrum, as well as its DRS/absorption spectrum for the photocatalytic system

under test, is essential in order to identify if it is truly photocatalytic in nature.

Acknowledgements:

SKL wishes to thank the DGIST R&D Program of Ministry of Science, ICT and Future Planning of Korea (17-NT-02) for supporting this work.

38

References

(1) P. Pichat, Photocatalysis and water purification from fundamentals to recent applications, 1st edn, Wiley-VCH, Germany, 2013 and references therein.

(2) K. Rajeshwar, M. E. Osugi, W. Chanmanee, C. R. Chenthamarakshan, M.V. B. Zanoni, P. Kajitvichyanukul and R. Krishnan-Ayer, J. Photochem. Photobiol., C, 2008, 9, 171–192.

(3) B. Ohtani, J. Photochem. Photobiol. C., 2010, 11, 157-178 and references therein.

(4) G. Zhang, G. Kima and W. Choi, Energy Environ. Sci., 2014, 7, 954–966, and references therein.

(5) T. Wu, G. Liu, J. Zhao, H. Hidaka and N. Serpone, J. Phys. Chem. B, 1998, 102, 5845–5851.

(6) A. Mills, N. Wells and C. O’Rourke, J. Photochem. Photobiol., A, 2017, 338, 123–133.

(7) B. Ohtani, Chem. Lett., 2008, 37(3), 216–229.

(8) A.V. Emeline, A. Salinaro and N. Serpone, J. Phys. Chem. B, 2000, 104, 11202–11210 and references therein

(9) T. Torimoto, Y. Aburakawa, Y. Kawahara, S. Ikeda and B. Ohtani, Chem. Phys. Lett., 2004, 392, 220–224.

(10) A.V. Emerline, X. Zhang, M. Jin, T. Murakami and A. Fujisima, J. Phys. Chem. B, 2006,

110, 7409-7413 and references therein.

(11) X.L. Yan, T. Ohno, K. Nishijima, R. Abe and B. Ohtani, Chem. Phys. Lett. 2006, 429, 606–610.

(12) R. Quesada-Cabrera, A. Mills and C. O’Rourke, Appl. Catal. B, 2014, 150–151, 338–344.

(13) T. Torimoto, N. Nakamura, S. Ikedaab and B. Ohtani, Phys. Chem. Chem. Phys., 2002, 4, 5910–5914.

(14) A.V. Emeline and N. Serpone, J. Phys. Chem. B, 2002, 106, 12221–12226.

(15) U. Stafford, K.A Gray, P.V. Kamat, J. Catal., 1997, 167, 25–32.

(16) K. Yamaguti and S. Sato, J. Chem. Soc., Faraday Trans. 1, 1985, 81, 1237–1246.

(17) A. Mills and G. Porter, J. Chem. Soc., Faraday Trans. 1, 1982, 78, 3659–3669.

(18) M.A. Grela, M.A. Brusa and A.J. Colussi, J. Phys. Chem. B, 1997, 101, 10986–10989.

(19) R. Abe, H. Takami, N. Murakami and B. Ohtani, J. Am. Chem. Soc., 2008, 130(25), 7780–7781.

(20) K. Nishijima, B. Ohtani, X. Yan, T. Kamai, T. Chiyoya, T. Tsubota, N. Murakami and T. Ohno, Chem. Phys., 2007, 339, 64–72.

(21) A.V. Emeline, X. Zhang, M. Jin, T. Murakami and A. Fujishima, J. Photochem. Photobiol., A, 2009, 207, 13-19.

39

(22) E. Kowalska, R. Abe and B. Ohtani, Chem. Commun., 2009, 241–243.

(23) J. Liu, Y. Liu, N. Liu, Y. Han, X. Zhang, H. Huang, Y. Lifshitz, S-T. Lee, J. Zhong and Z. Kang, Science, 2015, 347, 970–974.

(24) X. Yan, T. Ohno, K. Nishijima, R. Abe and B. Ohtani, Chem. Phys. Lett., 2006, 429, 606–610.

(25) T. Takizawa, T. Watanabe and K. Honda, J. Phys. Chem., 1978, 82, 1391–1396.

(26) J.R. Darwent, J. Chem. Soc., Faraday Trans. 2, 1981, 77, 1703–1709.

(27) J. Torrent, V. Barron, Diffuse Reflectance Spectroscopy. In Methods of Soil Analysis Part 5—Mineralogical Methods, A.L. Ulery, L.R. Drees, Eds.; Soil Science Society of America, Inc., Wisconsin, 2008, ch. 13, 367–385.

(28) A.V. Emeline, G.N. Kuzmin, D. Purevdorj, V.K. Ryabchuk and N. Serpone, J. Phys. Chem. B, 2000, 104, 2989–2999.

(29) X. Yan, T. Ohno, K. Nishijima, R. Abe and B. Ohtani, Chem. Phys. Letts., 2006, 429, 606–610.

(30) M.I. Cabrera, O.M. Alfano and A.E. Cassano, J. Phys. Chem., 1996, 100(51), 20043–20050.

(31) M.L. Satuf, R.J. Brandi, A.E. Cassano and O.M. Alfano, Ind. Eng. Chem. Res., 2005, 44(17), 6643–6649.

(32) A. Mills, C. O'Rourke and K. Moore, J. Photochem. Photobiol., A, 2015, 310, 66–105, and references therein.

(33) T.A. Egerton, C.J. King, J. Oil Col. Chem. Assoc., 1979, 62, 386–391.

(34) K.Y. Okamoto, H. Yanamoto and H. Teneka and A. Itaya, Bull. Chem. Soc. Jpn. 1985, 58, 2023–2028.

(35) G. Al-Sayyed, J-C. D’Oliveira, P. Pichat, J. Photochem. Photobiol., A, 1991, 58(1), 99–114.

(36) S.K. Lee, A. Mills and C. O'Rourke, J. Photochem. Photobiol., A, accepted for publication, ref no.: JPC 10645. http://dx.doi.org/10.1016/j.jphotochem.2017.05.018.

(37) A. Fijishima, X. Zhang and D.A. Tryk, Surf. Sci. Reports, 2008, 63, 515–582.

(38) Degussa Technical Bulletin Fine Particles, No. 80, 'Titanium Dioxide p25, Manufacture – properties – Applications, 1998, Degussa AG, Frankfurt.

(39) T. Watanabe, T. Takizawa and K. Honda, J. Phys. Chem., 1977, 81, 1845-1851.

(40) A. Mills and C. O’Rourke, Green, 2011, 1, 105–113.

(41) C. O’Rourke and A. Mills, J. Photochem. Photobiol., A, 2010, 216, 261–267.

(42) A. Mills, M. Sheik, C. O’Rourke and M. McFarlane, Appl. Catal., B, 2009, 89, 189–195.

(43) A. Mills and N. Wells, Chem. Soc. Rev. 2015, 44, 2849–2864.

40

(44) K. Nishijima, B. Ohtani, X. Yan, T. Kamaia, T. Chiyoyaa, T. Tsubotaa, N. Murakamia and T. Ohnoa, Chem. Phys., 2007, 339(1–3), 64–72.

(45) A.G. Agrios, K.A. Gray and E. Weitz, Langmuir, 2004, 20(14), 5911–5917.

(46) S. Higashimoto, T. Nishi, M. Yasukawa, M. Azuma, Y. Sakata and H. Kobayashi, J. Catal., 2015, 329, 286–290.

(47) S-K. Lee and A. Mills, Platinum Metals Rev., 2003, 47(2), 61–72.

(48) Y. Tian and T. Tatsuma, J. Am. Chem. Soc., 2005, 127(20), 7632–7637.

(49) R.S. Sonawane and M.K. Dongare, J. Mol. Catal. A, 2006, 243(1–2), 68–76.

(50) N. Serpone a nd A.Salerno, Pure & Appld Chem., 1999, 71, 303-320.